Perplexity: A single metric can determine who will win the AI race
```
The ultimate key to victory in AI competition may depend on a single number.
Aravind Srinivas, CEO of AI search star Perplexity, said in a CNBC interview on Wednesday that the company able to achieve the highest "value per watt per user token" will command the highest valuation in the future.
He emphasized that this metric integrates accuracy, latency, cost, privacy, and intelligence level, making it the core standard for measuring the long-term competitiveness of AI companies. Meanwhile, Perplexity announced that its Personal Computer product will be available on Microsoft's Windows operating system, further expanding its intelligent AI coverage.
Srinivas’s remarks directly shift the market’s valuation logic for AI companies towards the ratio of energy efficiency to economic output, rather than pure revenue scale. Against the backdrop of rising investments in AI infrastructure, this perspective is of important reference value for investors evaluating related targets.
Energy Efficiency Ratio: Redefining AI Valuation Logic
The core metric Srinivas proposed is "value per watt per user token." He explained that the token is the basic unit of data processing for AI models. When users issue instructions to AI, tasks are broken down into several tokens, and each token requires energy to process. Thus, the company that can create the greatest economic value per unit of energy consumption will occupy the most advantageous competitive position.
"The company that can maximize this goal—balancing accuracy, latency, cost, privacy, and intelligence—is the long-term winner," Srinivas said.
He also questioned the high-revenue model of some current model providers. "Some model providers appear to make a lot of money because their models are very expensive... but that's just short-term revenue growth." This statement implies that revenue expansion driven by high pricing does not necessarily translate into lasting competitive advantage.
Agent Deployment: From Search to Complex Task Execution
Perplexity is accelerating its transformation into the AI agent field. In February this year, the company launched Perplexity Computer, positioned as an AI agent able to execute complex tasks over a long period, surpassing traditional Q&A style interaction.
On Wednesday, Perplexity announced its Personal Computer product will support Microsoft Windows, enabling the AI to access applications such as Word, Outlook, and local files on user devices. Previously, the product was available on Apple Mac.
This expansion marks Perplexity's extension of AI capabilities from cloud-based search to local devices and productivity tools, echoing the trend described by Srinivas as "the data center moving into your laptop."
Competitive Landscape: Huge Valuation Gap, Rising Pressure
Despite Perplexity's aggressive strategic narrative, its valuation gap with top competitors remains significant. Reportedly, Perplexity's latest valuation is about $20 billion, while Anthropic approaches $1 trillion, and OpenAI slightly surpasses $850 billion.
Meanwhile, OpenAI, Anthropic, and Google are all accelerating their layout in the AI agent space, forming direct competition with Perplexity. In this arena, the resource advantages of giants cannot be overlooked.
The energy efficiency metric raised by Srinivas may also be Perplexity's way of building a differentiated narrative for its own competitive path—seeking breakthroughs through more precise value creation efficiency, rather than mere computing power stacking, especially when its resource scale does not match its rivals.
Risk Warning and DisclaimerThe market involves risks, and investment must be approached cautiously. This article does not constitute personal investment advice and does not consider the specific investment goals, financial situation, or needs of individual users. Users should assess whether any opinions, views, or conclusions presented herein fit their particular situation. Investing based on this is at your own risk. ```